Ameliorating Patient-Caregiver Stigma in Early-Stage
Parkinson’s Disease using Robot co-Mediators
Ronald C. Arkin
Abstract. Facial masking in early stage Parkinson’s disease
leads to a well-documented deterioration (stigmatization) in the
patient-caregiver relationship. This research described in this
paper is concerned with preserving dignity in that bond where
otherwise it might be lost, through the use of a robot co-mediator
that will be capable of monitoring the human-human relationship
for lack of congruence in the perceived emotional states of the
parties concerned. This paper describes the component
architectural modules that are being used in support of this 5year effort, including an ethical architecture developed earlier for
the military and previous research on affective companion robots
for Sony and Samsung that are able to express affective state
through kinesics and proxemics.
1 INTRODUCTION
Parkinson’s disease affects seven million people worldwide [1]
(Fig. 1). An unfortunate side effect of the illness, due to the
occurrence of facial masking [2,3] in the patient, is a
deterioration (stigmatization) in the relationship between the
caregiver and the client [4,5] whether this occurs in a clinical
setting or at home. The net effect is a loss of dignity with respect
to the client due to their inability to adequately communicate
their internal affective state to the individual in charge of the
person maintaining their care. Although spoken language (with
some loss of prosody) still is available, most of the nonverbal
subconscious cues associated with the client’s emotional state
are lost. This ultimately affects the quality of care given the
patient with repercussions in both economic and human rights
considerations.
Figure 1: Global Distribution of Parkinson’s Disease [1]
In a new 5-year National Science Foundation National
Robotics Initiative Grant, the Georgia Institute of Technology
and Tufts University1 are exploring ways to prevent a decline in
these otherwise stigmatized human-human relationships through
the introduction of a robot co-mediator. The goal is to enable this
robot to provide subtle nonverbal emotional cues that are
otherwise unavailable from the patient, in an unobtrusive manner
using kinesics and proxemics. Our previous research with Sony’s
QRIO robot in this area [6,7] provides the basis for an
implementation in a small humanoid robot using at a minimum
an Aldebaran Nao and eventually a Robokind R25 humanoid
platform. We anticipate investigating other robotic morphologies
as well.
We draw on our earlier computational models of moral
emotions [8], initially deployed in research for modelling guilt in
autonomous robots [9]. This research now expands the available
affective repertoire to capture empathy levels for the caregiver
while simultaneously modelling shame and embarrassment in the
client. The goal is not to “fix” the relationship between the two,
but rather to achieve congruence between the caregiver and
patient through mediation, so that each human has a true
representation of what the other is experiencing, This lack of
congruence is symptomatic of the resulting stigma. While there
has been some significant research to date in developing
mechanisms to deploy “empathy” to a robot (e.g., [10]), it
remains a largely unsolved problem. Here instead we are trying
to develop effective theory of mind models (or partner modelling
[11,12,13]) that capture and represent the levels of the associated
human’s moral emotions so that the robot can, in a homeostatic
manner, then drive them back to acceptable levels if either
empathy, as observed in the caregiver, is suffering from a
shortfall, or to provide appropriate feedback to the caregiver
should shame or embarrassment in the client reach unacceptable
levels.
From a systems level to ensure these requirements are met,
we are leveraging our previous research on ethical architectures,
specifically the ethical adaptor and the ethical governor [12,14],
to provide the computational representations and algorithms that
ensure these ethical boundaries are suitably enforced. Sensing
information will be derived from physiological sensors (e.g.,
galvanic skin response, FNARS, blood pressure, heart rate, etc.)
and direct signalling through a user-manipulated paddle. Parsing
of the spoken language of the client, where feasible, will also be
[1] School of Interactive Computing, Georgia Institute of Technology,
Atlanta, GA, U.S.A. 30332. Email: arkin@gatech.edu. This research is
supported by the National Science Foundation under grant #IIS-1317214
and is conducted in collaboration with Matthias Scheutz (Computer
Science Department) and Linda Tickle-Degnen (School of Occupational
Therapy) at Tufts University.
evaluated, but this time with believability attached to it, which
unfortunately can be lacking in the caregiver under these
circumstances.
This paper is organized as follows, focusing on the supporting
technology that this new research builds upon. In Section 2 we
review the aspects of the ethical architecture including the
ethical governor and ethical adaptor components and their
relevance to this research. In Section 3 we describe the role of
kinesics and proxemics, drawing on our earlier research for Sony
[6,7] and Samsung [15] in social robots and how this can afford
the expressiveness needed for this healthcare application. The
final section concludes the paper.
Emotions. These affective states vary in terms of their specificity
and spatial extent (Fig 4).
2 AN ETHICAL ARCHITECTURE AND
MORAL EMOTIONS
Figure 3. Cognitive Model for Guilt [8]
We now summarize the relevant components of our previous
work and how they can be integrated and extended into this new
healthcare application. Figure 1 illustrates the ethical
architecture developed in earlier research for the U.S. Military
[16]. The specific components being leveraged from this work
include the ethical adaptor (that manages emotional restriction of
action among other things) and the ethical governor (that
restricts the robot’s behaviour to acceptable ethical actions).
Here we focus on the adaptor as that houses the component for
managing the moral emotions, which we will tie into more
sophisticated means of controlling robotic emotions that are
described in Section 3.
Figure 4. Relationship between various affective forms
The combined ethical architecture and TAME (Fig. 5) are being
housed within our mission specification software system,
MissionLab [17] which provides the infrastructure for this
research.
Figure 5. TAME Implementation within MissionLab
Figure 2. Ethical Architecture [16]
Figure 3 shows the basis of our first foray into moral emotions,
specifically guilt [8]. This equation, based on cognitive science,
provided the basis for implementation in the ethical adaptor. We
anticipate using variants of this formula to capture other relevant
moral emotions such as empathy, embarrassment, and shame.
The next major component we apply is the TAME
architecture [15], which is one of the most sophisticated
affective control architectures to date. TAME encompasses the 4
primary affective phenomena: Traits, Attitudes, Moods, and
The net result is the ability of a small humanoid robot (Nao) to
express through nonverbal communication (kinesics in this case)
its internal affective state (Fig. 6 and 7).
Figure 6. Examples of affective expression in Nao robot
Figure 9. Various nonverbal cues based on human kinesics
Figure 7. Varying Displays of Intensity of Robotic Moods
achieved via TAME
Details of the TAME architecture can be found in [15,18,19].
3 ROBOT EXPRESSION VIA KINESICS AND
PROXEMICS
Our earlier research for Sony Corporation in the context of
companion robots led to the development of a behavioural
overlay architecture that provides mechanisms for expressing
nonverbal communication on top of on-going intentional
behaviours [6] (Fig. 8), and demonstrated on the QRIO
humanoid (Fig. 9). Specifically we were concerned with
kinesics (body language based on postures, emblems and other
non-verbal signals) and proxemics (maintaining an appropriate
spatial separation between user and robot). The proxemics
component also builds upon our earlier work [20] (Fig. 10) based
on psychological attachment theory [21]. It is our intent to use
these methods and mechanisms as appropriate to convey the
internal state of the patient to the caregiver in a subtle nonintrusive manner when needed as determined by a lack of
congruence observed between the theory of mind models
maintained in the robot of both human actors.
Figure 8. Behavioural overlay mechanisms for generating
nonverbal communication (See [6] for details)
Figure 10. Agent-Based Responsive Environments
4 CONCLUSIONS AND FUTURE WORK
As stated earlier, this is a collaborative 5-year project with Tufts
University just beyond its beginnings with much more work
remaining to be performed. At this point we are integrating
multiple components from our previous research including the
ethical architecture, TAME, and mechanisms for generating
proxemics and kinesics, all in order to faithfully express the
internal state of the patient to the caregiver when a lack of
congruence is observed. The intention is to reduce the stigma
[4,5] in the patient-caregiver relationship that arises due to facial
masking [2,3] in an early stage Parkinson’s patient.
The research plan for the Georgia Tech effort in the years
ahead is as follows:
•
Year 1: Characterization of Theory of Mind Models for Use
in the Architectural Framework
•
Year 2: Implementation and Testing of Positive Moral
Emotions in Ethical Adaptor
•
Year 3-4: Integration of Ethical Adaptor and Ethical
Governor in Tufts Testbed
•
Year 4-5: Collaboration with Evaluation at Tufts and
Iterative Refinement of Ethical Interactions
We expect that there will exist the possibility of generalization
of this research into other healthcare applications, perhaps
wherever concern exists between an on-going relationship
between client and caregiver. We focus for now, however, on
the specifics of early stage Parkinson’s disease in the hopes of
improving the quality of care and thus the quality of life for the
millions unfortunately afflicted with this illness.
REFERENCES
[1]
Wikipedia,
“Parkinson’s
Disease”,
accessed
1/24/14,
http://en.wikipedia.org/wiki/Parkinsons.
[2] K. Takahashi, L. Tickle-Degnen, W. J. Coster, and N. Latham.
Expressive behavior in Parkinson’s disease as a function of interview
context. American Journal of Occupational Therapy, 58:603–614,
2004.
[3] L. Tickle-Degnen, L. A. Zebrowitz, and H. Ma. Culture, gender and
health care stigma: Practitioners’ response to facial masking
experienced by people with Parkinson’s disease. Social Science and
Medicine, 73:95–102, 2011
[4] D. Rao, S. W. Choi, D. Victorson, R. Bode, A. Peterman, A.
Heinemann, and D. Cella. Measuring stigma across neurological
conditions: The development of the stigma scale for chronic illness
(ssci). Quality of Life Research, 18(5): 585–595, 2009.
[5] L. Tickle-Degnen, L. A. Zebrowitz, and H. Ma. Culture, gender and
health care stigma: Practitioners’ response to facial masking
experienced by people with Parkinson’s disease. Social Science and
Medicine, 73:95–102, 2011.
[6] Brooks, A. and Arkin, R.C., "Behavioral Overlays for Non-Verbal
Communication Expression on a Humanoid Robot', Autonomous
Robots, Vol. 22, No.1, pp. 55-75, Jan. 2007.
[7] R. C. Arkin, M. Fujita, T. Takagi, and R. Hasegawa. An ethological
and emotional basis for human-robot interaction. Robotics and
Autonomous Systems, 42(3–4), 2003.
[8] D. Smits and P. D. Boeck. A componential irt model for guilt.
Multivariate Behavioral Research, 38(2):161–188, 2003.
[9] P. Ulam and R. C. Arkin. An ethical adaptor: Behavioral
modification derived from moral emotions. IEEE International
Symposium on Computational Intelligence in Robotics and
Automation (CIRA-09), Daejeon, KR, December 2009.
[10] A. Tapus and M. Mataric, “Emulating Empathy in Socially
Assistive Robots”, AAAI 2006.
[11] A. Wagner and R. Arkin. Recognizing situations that demand trust.
In Proceedings of the 20th international Symposium on Robot and
Human Interactive Communication (Ro-Man 2011), Atlanta, August
2011.
[12] R. C. Arkin, P. Ulam, and A. Wagner. Moral decision-making in
autonomous systems: Enforcement, moral emotions, dignity, trust
and deception. Proceedings of the IEEE, 2012.
[13] G. Briggs and M. Scheutz. Facilitating mental modeling in
collaborative human-robot interaction through adverbial cues. In
Proceedings of the SIGDIAL 2011 Conference, pages 239–247,
Portland, Oregon, June 2011.
[14] R.C. Arkin, "Governing Lethal Behavior: Embedding Ethics in a
Hybrid Deliberative/Reactive Robot Architecture - Part II:
Formalization for Ethical Control", Proc. 1st Conference on Artificial
General Intelligence, Memphis, TN, March 2008.
[15] L. Moshkina, S. Park, R.C. Arkin, J.K. Lee, H. Jung, "TAME:
Time-Varying Affective Response for Humanoid Robots",
International Journal of Social Robotics, 2011.
[16] R.C. Arkin, Governing Lethal Behavior in Autonomous Robots,
Taylor-Francis, 2009.
[17] D. MacKenzie, R.C. Arkin and J. Cameron, "Multiagent Mission
Specification and Execution", Autonomous Robots, Vol. 4, No. 1,
Jan. 1997, pp. 29-57.
[18] S. Park, L. Moshkina and R.C. Arkin, "Mood as an Affective
Component for Robotic Behavior with Continuous Adaptation via
Learning Momentum", Proc. 10th IEEE-RAS international
Conference on Humanoid Robots (Humanoids 2010), Nashville, TN,
Dec. 2010.
[19] S. Park, L. Moshkina, L. and R.C. Arkin, "Recognizing Nonverbal
Affective Behavior in Humanoid Robots", Proc. 11th Intelligent
Autonomous Systems Conference, Ottawa, CA, Aug. 2010.
[20] M. Likhachev, and R.C. Arkin, "Robotic Comfort Zones", Sensor
Fusion and Decentralized Control in Robotic Systems III, Boston,
MA, November 2000, pp. 27-41.